Arao AI
Back to search results

gemma-4-E2B-it · Q3_K_L

Start a discussion
NVIDIA RTX 509032GBllama.cpp v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E2B-it (Q3_K_L) on NVIDIA RTX 5090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

324.42 tok/s

Decode

Decode speed

1,387.31 tok/s

Prefill

Prefill speed

0.22 s

TTFT

Time to first token

5.12 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. llama.cpp · Q3_K_L (current)Decode 324.42 · Prefill 1387.31 ·
  2. Strata · IQ3_S Decode 332.29 · Prefill 1223.54 ·

Core figures

Decode (avg)
324.42 tok/s
Prefill (avg)
1,387.31 tok/s
TTFT (avg)
0.22 s
VRAM (avg)
5.12 GB
MTP acceptance rate
—
TTFB
2.6 GB
Power draw
274.1 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E2B-it
Quantization
Q3_K_L
Framework
llama.cpp
Version
v0.53
Context length
4096 tokens
Batch size
256
GPU layers
28
Flash Attention
Off

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA RTX 5090
Nominal VRAM
32 GB
Measured VRAM (avg)
5.12 GB
OS
Ubuntu 22.04
Driver
565.57
CUDA
12.6
Power draw
274.1 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.53 · Ubuntu 22.04 · CUDA 12.6 · 4096 ctx

    324.42 tok/s

    Decode

    1,387.31 tok/s

    Prefill

    0.22 s

    TTFT

    5.12 GB

    VRAM

    —

    MTP

    274.1 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=13 running model_id=66 at Q3_K_L